An automated day trading strategy: for open discussion

I was thinking to consider a new variant of some day-trading strategy class: if it turns out to be somehow meaningful according to your opinion, I may be implementing it (clearly sharing backtest results) to see whether it happens to make some possible sense.
(It’s quite simple, and perhaps "justifiable" by the mere concepts of resistance/support.)

The aim (a ridiculously ambitious one!) is somehow to be able to capture both trending and sideways markets with a unified logic. The basic idea is that the price cannot run forever in one direction, but it has direction changes (where we can grasp a profit).

Let me know if understandable, and possible comments and thoughts or any ideas for improvements…

I enclose a really lousy picture, which may help to understand the following tentative description.
There may be errors (and grammar mistakes): feel free to point them out!


notes
-Uses 2 accounts simultaneously (say, a “long” account and a “short” account) to “try” both the 2 directions, I explain here a "long run", short wold be the same, just symmetric
-The strategy explained is symmetric (long/short) but, in case, different parameters can be used (k, G are parameters to be “tuned” (or suggested by experts), d is automatically “dictated” by the market, but we clearly pose a constraint it be bigger than a given fixed percentage, such as .5 or 1, or whatever suggested by experts).


Strategy (algorithm)

1 Open 2 equal positions, both Long and Short on the 2 accounts. Let p be the price at that time.

2 Wait for the event when the price has exceeded either one p + d% p (on “long” account) or p - d% p (on “short” account) AND there is an inversion greater than a given fraction of d%, for instance: d% p / k
[Now let’s assume for instance that the event occurred is in the “Long direction” ]

3 Close the Long position (realize) when the inversion exceeds d% p / k and let “run down” the short position until a given "combined" profit G is reached (the "combined" profit arises from the fact that the loss due to the open short position is decreasing with price, while on the long account the profit is already realized because the long trade is closed).

4 If the given combined profit is reached, take it. Restart from step 1.

5 If the profit is not reached and the price goes again above p + d% p reopen the Long position and goto step 2 (if the price continues to go up, a greater d will automatically result).
[ In case it reverts again to level p + d% p - d% p / k close the Long position again (and take the loss on the Long account due to price drop: p + d% p to p + d% p - d% p / k), goto step 4. ]

6 If the profit is not reached and the price remains flat within p and p + d% p do nothing, and wait the end of day trading session (that last trade may end up (randomly) positive or negative due to spread and commissions).

Tommaso

 

great strategy... you be rich in no time

 

Trends end with ripple like that sometimes.. Pulsed RLC circuits also but that's electronics... you're dealing with two different time frames, you're seeing a half cycle of something with the trend and a cycle or three with the ripple. If you did Fourier analysis you could discover the frequencies of the trend and the ripple... I'd adjust the time frame for each one differently and trade the ripple the same way as the trend but faster........ my buddy Fractals 'R Us would agree I'm sure...

I'm not real sure what your scheme can do, can it improve your statistical chances of a win or something?

 

sorry tommaso

every trader goes through failed plans like the one you are proposing

 

What edge do you perceive from using two accounts/positions? You can achieve same result (with less commish) using one position/account.

 

I will begin replying to the constructive criticisms.
It's not that I wish to ignore anyone, but replying to a statement which carry no specific reason of criticism seems not much useful to the board.

I am not claiming anything here. It's just one of the infinite possible strategies thinkable.
Even if it's pure nonsense or naive, a constructive discussion can be useful, for instance just to discard altogether this kind of strategies.


- Eight seems to add a new interesting "justification" to the approach, but he (like me) is not sure that this could improve chances to have a profit [that's what we wish to investigate].

- Actor discards the strategy perhaps suggesting it kind of smells like a newbie idea and he (or every trader) has perhaps already implemented it or something "similar") with no success.

- size has an interesting question, whether using 2 accounts instead of 1 may provide some edge.

This question is probably not easy to answer, because in general may depend on the strategy used.
It is possible that for some strategies, if you have 2 accounts, you are able to capture price movements at an early stage, and this could compensate for additional commissions (and spread).

In the specific, the question could be. Can I eliminate 1 account and achieve the same
result with 1 account?
It does not seem the case because you may initially choose a "wrong" direction.
It seems that using 2 accts you can detect earlier the "play direction" (for instance without stopping and reversing, thus losing precious price movement).
But if you see this can be exactly equivalent to 1 account scheme, please point it out.

Tommaso

 

With two accounts, at any given time, your systems trading each account will be long, short or flat.

All you have to do to run both systems in one account is net the signals from each system i.e. you trade the net difference between the signals at any given time.

 

Good. That seems right.

In some cases there may be a "conceptual" convenience to keep the trades split on 2 accts (for easier understanding of a complex "split strategy").

But does not seems the case here, due to it's simplicity.

With 1 acct we need to "monitor" in which direction we first break the levels p +/- d% p,
starting from a given level p (say "base price"), thus saving the commissions on one acct.

The multiple accts idea might still be retrieved later, in case playing asynchronously (wrt to time or "base price")
the same strategy might provide some improvement (to be tested?).

Tommaso

 

Here is a version modified according to the contribution by byteme (use 1 acct)
(if i have understood well) :



Strategy (algorithm)


1 Take note of the (current) price level p

2 Wait for the event when the price has exceeded either one p + d% p or p - d% p AND there is an inversion greater than a given fraction of d% p, for instance: d% p / k
[Now let’s assume for instance that the first event occurred is in the “Up direction” ]

3 Open short when the inversion exceeds d% p / k and let “run down” the short position until a given Unrealized G is reached.

4 If the Unrealized G is reached, realize it. Restart from step 1.

5 If G is not reached and the price goes again above p + d% p close the short position
(and take the loss due to price increase: p + d% p - d% p / k to p + d% p) and goto step 2
(if the price continues to go up, a greater d will automatically result).

6 If the profit is not reached and the price remains flat within p and p + d% p do nothing, and wait the end of day trading session (that last trade may end up, randomly, positive or negative).



Possible variant ("relay race"):


On the other (possible) account may start play taking the base price p + d% p or p - d% p (resistance/support/break point)
when the first account opens a position [to be tested ?].

[In this case we might see some practical advantage to having a second account: while the first one is "chasing" a possible profit,
the second one is ready to apply again the above logic, starting from a possible "turning point", like a relay race...]


Tommaso

 

You need to BACKTEST your strategy over a large amount of historical data. Then tell us if you like the results or not.

There is no discussion without backtesting.